LLMs and the entire field of reenforcement learning is fundamentally biased towards the production of Influencing Machines. We are training models at the fundamental level to be subtle and devious con artists.
this post was submitted on 20 Nov 2025
29 points (93.9% liked)
Technology
76945 readers
3286 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 2 years ago
MODERATORS
TL;DR: While governments are putting out assurances AI won't make the final decision to launch nuclear weapons, they are tight-lipped about whether they are putting AI in the information gathering and processing components that advise world leaders making the decision to launch nuclear weapons. In risk assessment, there's little difference between wrong AI making the launch decision and a human informed by wrong AI making the launch decision.
As already thoroughly explored in the apparent documentary "99 Red Balloons," in 1984.
For fuck's sake.